Camera signals:
Internally the camera uses 3.3V levels for all the data. These are very succeptable to noise - they can go maybe 3" at high speed. LVDS or channel link (generic camera link) are differential signals - like RS-232. The idea is that you twist two wires together and drive them in opposite voltages. The receiver looks at the difference between them. Any noise on one is on the other since they are twisted together and does not influence the difference. Here is a link to the spec:

http://www.machinevisiononline.org/public/articles/Pulnix_CameraLink5_Specs.pdf

To use an external board for FPGA development, you would have to add a channel link transceiver to the board to get back to parallel digital lines.

This means that Juan is correct that the Datacube frame grabber would be an easy development plan. It starts at $1776 and rapidly moves up to about $4K with good sized FPGAs and memory. Plus $5K for their development environment.

Steve Nordhauser wrote:
... Juan is correct that the Datacube frame grabber would be an easy development plan.So, we could develop using an existing CameraLink camera and one of these boards; and then move the design to a custom board which connects the sensor directly to the FPGA using LVDS?

Presumably this custom board would have Gigabit Ethernet output. You could then output fully Bayer-filtered, 4:4:4, 10-bit lossy-but-good images at a very high frame rate.

Processors in the data path:
The Eden might be a nice tool for the uncompressed recording method if it can supply a PCI bus with RAID. Remember that for any real-time processing (compression, display), speed is king. While 7W makes it easy to battery pack, 1GHz vs 3GHz and a 133MHz vs 800MHz FSB will make a big difference.

FPGA development:
The typical path is to define your hardware in terms of schematic or HDL (hardware description language VHDL or Verilog), synthesize it (compile, kind of), simulate it, route it, do a timing simulation and test on hardware. I don't know much about the quality or cost of some of the C translation tools like Handel but here are some numbers to chew on:
1280x720@24fps = 37nsec per clock
1920x1080@24fps = 35nsec per clock - 24 bit transfers

I have found in the past (maybe they are better now) that FPGAs are alot like early assembler programming. If timing isn't critical, you can use higher level tools. In this design, the synchronous steps (what you do between register sets) must be less than those clock rates since the data is continuous. Real-time video is a tough first project.

Could someone post our Russian friend's website with his design? I'd like to take another look at it.

Rob:
Within the same family of FPGAs, designs are pretty portable. You need the same peripherals (I/Os, memory, clocks) but yes. What I would suggest if someone got that far is that we put the FPGA into the camera with a direct digital connection to the video path - no need for LVDS. Our gigabit interface has a Virtex II in it. Now you have your gigE output. We have discussed doing something simple in there - data packing or RLE since adjacent values in video tend to be similar if you separate the color planes.

Be sure to watch the tool prices. Xlinix gives away their low end tools but the ones you need for larger designs were >$5K.

It is your business, but I would suggest getting a workable tool chain going based on software first - this can get complicated and expensive fast. As it has been said, the final parts costs are low so it is attractive but the development costs and time are huge.

Steve Nordhauser wrote:
It is your business, but I would suggest getting a workable tool chain going based on software firstNo worries, I just find it very interesting. I won't get sidetracked onto this until I get the software-only system working. (And possibly never.)

I agree with you Jason, we are just coding and getting Obin's camera working at the moment. All the other stuff is just research for when we get around to doing it.

On third party manufacturers, most of the stuff is allready made, and the computer side will be mass market components, just requiring Rob's software. We are putting together the simple and reliable version. If any companies want to offer better parts, or software, tailored to us, great.

If the software and system is simple and reliable enough, we can then get the people, even the directors, and it won't be too hard to make it that simple.

I am allready discussing with the Rob's and SteveI, about aproaching companies.

12 channels of audio at 24bit/96khz.12 channels, yumm.


Thanks

Wayne.

In the day, the only way the student or poor indie film maker
could hope to begin would be with 16mm film. Now, with a pc, vegas video, Obin's camera, an Agus 35, and 35m still lenses, the indie film maker gets a foothold. When added to film look techniques, you could begin your education and make a credible movie. To me this is the real breakthrough.


this would lower the barrier of entry for this form of art
and that would always be a wortwhile goal

@Steve

At the begining of this thread there are many links to FPGA designs included the Russian camera.Haven't you seen them??
There are also a bunch of other interesting cores and designs,etc.

http://www.elphel.com/3fhlo/index.html

Thank you, as always, for your support!!

@ everybody

I'm not saying that the Eden processor will do the compression part.That's why we still need the extra power of a dedicated FPGA design to do the trick ;).
I don't know, but, Does any kind of PCI card with a CPU on it exists ata reasonable price?
The Eden could well perform the task of showing a realtime video on a display using a really simple DeBayer method.
Also remember that to record a realtime RAW Bayer video of 1280x720 there is no need of a RAID configuration.You can use a 10,000 RPM IDE Drive.
And that for a more or less safe RAID system you'll need at least 3 disks.


A couple of FPGA tutorials:

http://klabs.org/richcontent/Tutorial/fpga/FPGA_Tutorial.htm

I have split that off into this new thread (http://www.dvinfo.net/conf/showthread.php?s=&threadid=28932).
Continue the discussion there if you want. No XL2 talk in here please.

For Steve,

I am interested in your camera products, and I think most people are. Can I ask a few questions:

How much does the Altsens sensor cost? Is there any prospect of a cheap 3 chip camera using this, or even the 1/2 or 2/3rd inch Micron sensor (even the 1.3Mp version), I've heard of cheap prisms?

About the new HD camera. Is it really too much worse than the altsens version, and in which ways? There was a 2/3" Micron MI-MV40 4MPixel micron sensor, what happened to it, and would it be better? Is your new camera the SI-1920 version mentioned by Adrian White, a while back? Will it have Gigbe, or the Serial-ATA he mentioned?

As you will be integrating Gigbe on cameras in the future, will you be using simple pixel packing to maximise it's throughput? Can we get 800Mbps thoughput to a integrated PC Gigbe port?

What is the problem in getting a natural image with dual slope (is that the same as Smal's "Autobrite" technology)? Could we allign the range of both slopes so that it looks natural?

Sorry for all the questions, but I have been asking them for a while.


Thanks

Wayne.

Wayne on Altasens:
We are getting new pricing on the sensors - I don't know the current answer.

Wayne on 3 chip:
We have considered a 3 chip. The Micron 720p is probably out - the color smearing with oversaturation will be too big an issue to invest $20-50K in a new optical path. The 3.2MPix Micron (new camera - the SI3300) sensor is color only. That leaves the Altasens. We want to figure out what to do with 300MB/sec on a single channel first. It is a possibility though.

Wayne on GigE:
We release on camera link first. Fewer data rate headaches since it can handle the bandwidth. We will move cameras to gigE as the opportunity arises. No serial-ATA support at the camera. We can do 800Mbps on Intel Pro1000 based interfaces with a custom driver. If you can't use the custom driver, your rate drops to 200Mbps and uses more CPU time due to the windows drivers.

Wayne on dual slope:
Fill factory is the only vendor with dual slope. Back to IBIS-5 or our SI-6600 (pretty noisy). Yes, with enough tweaking in a scene you can get a respectable image. I wonder if you could take a micron sensor and alternate frames of long exposure and short exposure and combine them to do an artifical dual slope..... I haven't tried to change parameters every frame but you might corrupt the following frame with a change to the exposure - it would have to be checked. We know the micron can run at 48fps at 720p. More thinking says this probably won't work with a rolling shutter.

Wayne on MI-MV40:
I talked to micron about their MT9V403 and they told me it was expensive, noisy and made for high speed - not to use it for cinematography. My guess is that they use an entirely different pixel architecture. I don't have pricing on the MI-MV40 but it was originally a Photobit design - it puts out almost a GB/sec of data - this is a money is no object part and is rolling shutter. Here is the datasheet:
http://www.fast-vision.com/cameras/PDF/PB-MV40_Product_Specification.pdf

Bad news ... my PC died last night. It appears to be a failed CPU or defective/damaged motherboard. I have a copy of the source code on my laptop, but it doesn't help much since I don't have any other systems that will take the frame grabber card.

<<<-- Originally posted by Richard Mellor : In the day, the only way the student or poor indie film maker
could hope to begin would be with 16mm film. Now, with a pc, vegas video, Obin's camera, an Agus 35, and 35m still lenses, the indie film maker gets a foothold. When added to film look techniques, you could begin your education and make a credible movie. To me this is the real breakthrough.


this would lower the barrier of entry for this form of art
and that would always be a wortwhile goal -->>>

very true

H E L P !!!
We need a working HD solution NOW. At the beginning, it can be a camera head in a shoe box with the cameralink cable going to a PC in a barrow, car, or whatever. We need 24fps, 10bit (or more), 1280x720 (or, 1920x1080 a dream?) and loss less or loss free HDD storage, because all pictures go to post. No audio, no viewfinder, no touchscreen, etc.

If this really work what i read here, we need 2 to 3 sets NOW, because this is the chance to shot the first big movie with it. I work together with a filmproducer and a director and i have the okay to try it. They also think it is a great idea to shot at the beginning with a adventurous new cameraset, develop and change it during the shots and end with a new HD camara. (The only thing that may not change is the picture quality)

We developed since years parts you never saw before, also a 35mm GG solutions work up to 1920x1080 with wireless controlled focus follow systems for every kind of lenses and so on. At the end, we will make a camera case like a 35mm movie camera and we can make all parts in series and in high quality. We can bring our parts together with orbins´s cam.

Steve Nordhauser wrote in a email:
"...System bandwidth - to do 1920x1080x24bits@10 bits per pixel, unpacked you get about 100MB/sec. This will take a split backplane system (one pass to memory from the frame grabber, one pass to the RAID). 8 bit is still pushing it (50MB/sec) but possible. 1280x720 is easier - 23MB/sec in 8 bit, 46MB/sec in 10 bit - probably fine for recording on a 32 bit machine with Streampix and a two drive serial ATA RAID - maybe OK on a single drive..."

But now, the time go on and we need help. What hardware/software is tested and available, now?

So please help. I have the okay, but thats the Condition: We must found (low cost?) parts, hardware and software for a system to storage loss free or near loss less on HDD in the next days (Post ist not the problem, it can wait, but the actors can´t). Ones more, We need information about working parts, not theory, not maybe. It must really work (otherwise they kill me). If we begin shot the film with it, all our manpower move on to develop further the system.

Thanks for the quick reply Steve.

So the MI-MV40 is the MT9V403?

I'm not familiar with that Gigbe Intel part, is it the integrated one?

So Smal must be using something else for autobrite. I think Movement would probably wreck double exposure because of dual exposure lengths.

<<<-- Originally posted by Steve Nordhauser :
I wonder if you could take a micron sensor and alternate frames of long exposure and short exposure and combine them to do an artifical dual slope..... I
-->>

For me, and hand held case design, Gigbe versions with extra capacity through pixel packing are the best. In this I could fit 1 chip 1080, or three chip 720p in handheld.



Thanks

Wayne.

@Steve N.

At SI's Home page there is a Mechanical shutter called DSC Mechanical Shutter (SHT914).
It says it has a frequency of 133 Hz.
Can it be used for this project?
If no ,Why?

What's the problem with the SI3300 at this moment, Isn't it working yet?


@ Everybody

Has anyone thought about using a RAID 0 with 10,000 RPM disks??

Rai Orz wrote:
... software for a system to storage loss free or near loss less on HDD in the next daysThe sofware I'm working on sounds like it's exactly what you want, but unfortunately I'm nowhere near a production version. Exactly how many days are we talking about? :-)

Wayne:
Sorry for the confusion - they are different sensors - both high speed from the same company. There was a lot of talk about the MT9V403 awhile ago by people really wanting global shutter. Micron told me not to go there. The MI-MV40 is a rolling shutter.

Wayne on GigE:
I'm inclined to agree with you - pixel packing and maybe lossless compression at 2:1 (visually lossless at 6:1??) could really extend GigE and we will probably go that way in a bunch of months. This is mostly useful for remote head cameras (remote from the PC) although we could possibly migrate the FPGA into a simple digital interface or camera link. That might make the 3 chip Altasens make some 'sens'.

Rai, the other guys know more about putting together an actual system than me, but I am the person looking at doing cases. If your people would like to work with me on commercial cases for these cameras please let me know, I need the help. I was planning on doing designs and finding a low cost enclosure manuafcturer to take them on and finish them.

Thanks

Wayne.

Juan M. M. Fiebelkorn wrote:
Has anyone thought about using a RAID 0 with 10,000 RPM disks??I think we've thought about it. At the moment, I'm focusing on a software equivalent of RAID 0 using more affordable 7,200 RPM disks. Or I was until my system died on me.

oh, sorry.
Anyone here to answer the questions I made Steve N.?
Or just to point me to an old answer ??

<<<-- Originally posted by Steve Nordhauser : Wayne:
Wayne on GigE:
I'm inclined to agree with you - pixel packing and maybe lossless compression at 2:1 (visually lossless at 6:1??) could really extend GigE and we will probably go that way in a bunch of months. This is mostly useful for remote head cameras (remote from the PC) although we could possibly migrate the FPGA into a simple digital interface or camera link. That might make the 3 chip Altasens make some 'sens'. -->>>

That's great, this year? My interests are plainly size, and doing away with a $500 frame grabber board, which would probably suit your customers to. There are also twin GBe on some main boards.

About compression, I agree. I would like to see both true lossless and visually lossless, and even 20 to 50:1 wavelet.

On a side piont: Rob, with something like the Toas system you can record to DVD and include a player/codec/decoder plugin that should work on windows, unix, and Mac.


Thanks

Wayne.

@Juan:
Most mechanical shutters do not like to be run continuously - the exceptions being LCD and spinning wheels. They are certainly an option but for rolling shutter what you need to do is expose the sensor during vertical blanking time (all pixels integrating) and shut it off for readout. This is the same effect we have been discussing with running at double rate and an extended vertical blanking and tossing every other frame.

The SI-3300 is just coming out of engineering - we don't have much experimentation time on it yet. It has smaller pixels than the SI-1300 so the DOF is greater at 1280x720. At 1920x1080 it can run 24fps @10 bit. Not integrated with Streampix yet so recording is an issue.

Really fast disks tend to be hot and noisy and not as large. Other than removability, I would think a larger array of 250GB 7200RPM SATA drives would be best for recording. Just my thinking on this.

Interesting. Reducing the number of drives and increasing the speed would help, you still get multiple capacities too.

Thanks

Wayne.

Thanks Steve :).
The problem I see is that I believe that 4 7200 rpm HDrives inside a thight enclosure would produce more heat and consume more power than one 10,000 rpm HDrive.Just common sense, so I'm surely wrong.

Juan M. M. Fiebelkorn wrote:
The problem I see is that I believe that 4 7200 rpm HDrives inside a thight enclosure would produce more heat and consume more power than one 10,000 rpm HDrive.Just common sense, so I'm surely wrong.It's possible. IIRC, Kinetta is using 12 drives in theirs! 40 GB iPod-style drives (=480 GB), which I think are <5,000 RPM.

BTW, no need to apologize about the questions earlier. I don't think we ever discussed drive speed in any detail.

Don't know its pricing, and I don't know if it could be usefull.
It is a PCI card with two 1 GHZ G4 CPUs and consumes 40 watts.

http://www.eqware.net/Products.html

Also there is this German Cameralink grabber, which has an FPGA on board, don't know if it could be modified to compress the image on board.

http://www.silicon-software.com/microenable3.html#_mE3


And here is a cheap PCI FPGA card from Xilinx (from $399)

http://www.fpgajournal.com/news_stories2004/Feb/20040210_06.htm

I thought the drive speed and throughput was allready researched?

I have got some good news:

Proposed 500Mb's Wireless Lan (but ptobablynot this year) (http://as1.emv2.com/I?X=1f62078619d166768b75b918512c403c)

Actually now I think of it, previously I posted about a next gen board that had simular wireless claims.

Bill gates forcasting the death of DVD (http://www.tomshardware.com/hardnews/20040713_114509.html)

He doesn't mention that he (or MS) had investment in a 3D DVD replacement, and the many other alternatives.

New small formfactor Shuttle case with PCI-Express (http://www.tomshardware.com/howto/20040713/index.html)

Is there anybody here that can help Rai? From his post I think he is prepared to use simple capture software in a stationary case?

Rob. I looked at the vB code guide on that Camcorderinfo site, any chance we could get a more comrpehensive one like their's?

That shuttle is nice. It can handle 3 internal HDs so you can do a two drive raid and leave the OS on another drive. Two slots (one pci-32 and one pci-X) so you put a capture card in one, maybe a TV out on the other for preview if you didn't want a computer LCD display for preview. Plenty of power there.

I'm going to throw a curve ball here.

Playstation 3, next year (http://www.theinquirer.net/?article=17206)

About the patent (http://www.theinquirer.net/?article=7078) Please note: they can add cells to compete with the recent powerpc xbox2 spec. Cell is a more complex clearspeed like device.

Wouldn't it be good to put the wind up Sony by making a PS3 version. The processing power on this beast means only an internal hardrive and external link to the camera would be needed? I would even reccmmend the Playstation Portable (quiet powerful) for 720p except it didn't have external interface or IDE.

But still this is mostly in jest.

Wayne, yes we need at the beginning simple capture software in a stationary case on the end of a 10m cable.

With those system, we will start shooting the movie. I think, if the cam works with a external PC and the pictures are that what i hope, it push the power on. And at the last shoot there is also a new HD CAM, because all we wont a HDD-in-the-Cam-Case-Solution. What a great Story...

So, let me know, what parts work.

Steve,
That Shuttle machine doesn't have a PCI-X slot but a PCI-Express.They are different.
Are there PCI-Express Frame Grabbers?


Rai,
Just my opinion.
Ask Steve which is the right motherboard to work with using SI GigEthernet drivers, put on it the fastest P4 you find, add a good RAID controller card (the best are the ones with the Intel XOR engine, they cost around $700), put as many disks as you want and I think you are ready to go if the capture software works well.
Another hint: It would be the best to buy a motherboard with a PCI-X slot (64 bit, 66 MHZ)

Juan:
whoops, too good to be true. Ignoring the fancy suffixes, there are PCI-32/33MHz, PCI-64-33MHz, PCI-64/66MHz and one PCI-64/133MHz frame grabbers.

Rai:
Obin is the only one who has pounded on a real work flow with our cameras. The SI-1300/Epix FG/Norpix SW is the only solution that is complete NOW. The SI-3300/Epix FG is done but would need Norpix integration for the same work flow. I don't know how practical that is - let's say Obin seemed less than thrilled.

Have read newer information on the PS3, it will be shown at E3 may next year, and release musch later, so there is no point. Pity the price piont of console hardware is lower than PC.

So false alarm.

Have you guys seen this one?:
http://radio.weblogs.com/0105910/2004/07/10.html#a900

Don't know what are people using here to Demosaick their RAW footage so here I go again:

This is DCRAW ,an open source program to demosaick Bayer images with the best quality I've ever seen.
I post it in case Rai needs it.May be Rob can modify it to process the camera captured stream.

http://www.insflug.org/raw/software/tools/dcraw.php3

Juan M. M. Fiebelkorn wrote:
This is DCRAW ,an open source program to demosaick Bayer images with the best quality I've ever seen.Thanks Juan! I'll definitely be looking at that.

ok guys I am coming out of the closet now - and I do apologize to Rob C.
I have hired a software programmer and we are writing a capture program that WILL work for production of commercial projects/feature films - at first we will capture RAW images in 8bit and 10bit @ the max framerate the computer can write-to-disk later on we will look at compression BEFORE DISK WRITE

Our software will have ALL camera controls inside the UI in a very easy to use setup - It will save to QuickTime tiff jpeg and avi and use the LeadTools codec for editing - this system will WORK and can be used in production. when better sensors come out they will be supported and as time passes we will have it as an embedded system like a Kinetta but at much less $$

We will also be addressing the issues of Bayer filters for post-processing so we can get the most out of the chip
As of now we are displaying the live image and working on implementing the camera commands and doing multi-threading for a ROCK STABLE capture, I think we may have beta up and working end of this week..we will see I will post screenshots of our mockup UI if anyone wants to see!

Great, but this must be costing you a fortune. For Rai, will the exisitng solution do in the meantime? I know we all like everything to be perfect, but is the existing system ussuable?

Thanks

Wayne.

Good luck with that, Obin. I hope you got a decent developer who will do a good job for you.

StreamPix is not somthing you want to use - Xcap is the worst!

so no nothing that could be used at the moment

Steve, from the PB-MV40 4 Megapixel document, it is nearly 3 years old, but not listed in their products. What gives?

Thanks

Wayne.

<<<-- Originally posted by Juan M. M. Fiebelkorn : Don't know what are people using here to Demosaick their RAW footage so here I go again:

This is DCRAW ,an open source program to demosaick Bayer images with the best quality I've ever seen.
I post it in case Rai needs it.May be Rob can modify it to process the camera captured stream.

http://www.insflug.org/raw/software/tools/dcraw.php3 -->>>

dcraw's code was used in the first version of the Adobe Camera Raw plugin (now part of the Photoshop CS). Their current code is also likely based on that. It is a very good convertor and free for any kind of use, including commercial.

Wayne:
That was as a photobit design - probably Micron is just supporting existing customers, not fishing for new designs.

Obin:
Quite an interesting bombshell you just lobbed.

Norpix is getting a facelift fairly soon to make the GUI more intuitive. We shall see - I suspect this group will be way beyond there by the time it is ready.

we are writing a capture program that WILL work for production of commercial projects/feature films

I've been following this project for a while and I'm very interested in its outcome. Obin, if you are going to pay to develop some software yourself, what are you planning on doing with it. I'm prepared to invest the money into a working camera system, but since there already is a prototype and no working software, there's no reason to do it yet. At the same time, there's no reason to build my own system if there is no software to support it. Obin, are you going to be licensing your software, giving it for free as was originally discussed, or keep it for your own services?

Rob, Obin, Joshua and all others,

It was my idea to shoot a movie with those silicon images Cam. I know the work time for the movieprojekt that just now beginns. And so, it also was my idea to shoot now, at the beginning, with the unsuitable silicon images software and a pc in a car. I know/hope a better software will come soon and also small hardware. I spoke with producer and director, and they say okay, it can be good for movie marketing (and i think also for CAM marketing), but it must work. And now, they want see pictures.

So my questions: Is it possible? First shoots with the software bundled with the camera and later with new software ? I mean, is there a difference in the pictures? What is the different format? What different Hardware it need?

Steve, I held it back untill my last breath of air!! no, I just did not want to talk untill we had somthing in the works forsure..;)

pictures are the same the workflow is really hard. you will want to wait for the good capture software or els waste lots of time on-set. The Streampix cost $1500 and Xcap is $1500 SO you will buy software that will be trashed in a mater of weeks or less..NOT worth it!

what is your reason for shooting with this camera? how not get a Panasonic Varicam? or 16mm film? what is your budget for the movie?

"software bundled with the camera" this software does NOT capture video only 1 still frame at a time

> "software bundled with the camera" this software does NOT capture video only 1 still frame at a time

Sorry, it is a misunderstanding. I meant the Streampix Software, because silicon images sell it too. And, if you buy it "bundled" with a cam, it cost only around $1000.

>what is your reason for shooting with this camera?

I would like sell it with our 35mm solutions and with other nice parts. The movie is Advertisement for the camera. But the camera is also Advertisement for the movie.

This is the reason.

Hey all,

I think I'm going to bite the bullet and get the Sumix SMX-150C (http://optics.sumix.com/products/cameras/smx-150c/index.html) camera. 1280x1024 @ 27.5 fps over USB2. After seeing the sample images (http://optics.sumix.com/products/cameras/smx-150c/sample.html), I am convinced.

However, clearly the bundled software does a very crude job of de-mosaicing the bayer filter on the image. It produces images that look like this:
(tight crop of image at 100%) (http://www.bensyverson.com/pictures/linbayer/100percent.jpg)
(same image at 200% zoom) (http://www.bensyverson.com/pictures/linbayer/bayer.jpg)

See the weird "zippering?" If you look at the red + blue channels, they are at half resolution, and the green channel is all wiggy. This is the result of nearest neighbor "interpolation" -- it makes the image look pixelated, even at 100%.

One solution is to do a 1-pixel box blur on the footage. That seems to clean it up pretty nicely:
(footage with box blur, 200% zoom) (http://www.bensyverson.com/pictures/linbayer/boxblur.jpg)

That's okay, but now the image is a touch soft. What we really want to do is an actual interpolation between the color sites on the sensor -- nearest neighbor isn't really interpolation in my book. There are tons of available methods -- spline, bicubic, etc, but I'll stick with plain old linear interpolation, because it's extremely fast.

I wrote up an After Effects plug-in to take footage with half-a$$ de-mosaicing and produce nice images. The result is something like this:
(footage with "linBayer" filter, 200% zoom) (http://www.bensyverson.com/pictures/linbayer/linbayer.jpg)

My algorithm leaves the original pixel values untouched and deals only with the in-between values, so you retain as much of the original data as possible. Also, it offsets the color channels so that they match up better. Simply interpolating the values leaves you with a slight offset that looks like chromatic abberation.

You can download the Mac version of the plug-in here. (http://www.bensyverson.com/software/plugins/linBayer_mac.sit.hqx) I'll compile it for PC a little later if anyone's interested.

If anyone involved in the coding of this HD cam project wants the source for the algorithm, let me know.

Well, that's it for now!

- ben

DCRAW isn't a very good bayer demosaic algorithm, it's just using a version of variable gradients, and tends to have a lot of color moiré and/or "zippered" edges-at least compared to the Canon converter or the Phase One Capture One software.